Study of MFCC, Spectral Peaks, and Hashing for Real-Time Audio Fingerprinting in Quran Classification Systems

Oral recitation of the Holy Quran in millions and fractions of Muslims globally is a well-established tradition, which spans over generations. The culturally and religiously important recitations are widespread, making it an imperative to protect, label and identify them. The complexity and subtle differences between Quranic recitations, in terms of style, rhythm or even individual Qaris (reciters), make it difficult for existing classification systems to recognize them automatically. In light of these problems, this paper proposes a new classification of the Quran using an audio fingerprinting based Mel Frequency Cepstrum Coefficients (MFCC) with spectral peaks and hashing system. It is a system that helps to store and retrieve as well as classify recitations of the Quran, can make it easier known more effective fingerprint audio created from features MFCC. Designed to locate Surahs and Ayas within audio recordings with high sensitivity, recognizing a wide variety of recitation styles. We demonstrate that this approach is able to surpass algorithms on mutually exclusive features space endowed with MFCC (since it encodes rich characteristics) than those working only over simple spectral based models especially in the case of reciter and Tajweed discrimination. This shows the capability of this system in Quranic research aspect, educational purposes and especially for digital quaran repositories.